llms.txt

# ellmos-ai

Canonical organization: https://github.com/ellmos-ai
Organization profile repository: https://github.com/ellmos-ai/.github

ellmos-ai builds local-first infrastructure for LLM operating systems, agent memory, model routing, multi-agent coordination and Model Context Protocol servers.
The ecosystem is centered on Python 3.10+, SQLite and MIT-licensed tools with zero or minimal external dependencies.
Public repository coverage: this file lists all public ellmos-ai repositories known on 2026-08-05 (51 repos, 1 archived); archived repositories are marked.

## Primary entry points

- BACH: https://github.com/ellmos-ai/bach
  Full personal LLM operating system with GUI, memory, tools, skills and agents.
- Rinnsal: https://github.com/ellmos-ai/rinnsal
  Lightweight zero-dependency LLM infrastructure for memory, tasks, connectors, chains and i18n.
- ellmos: https://github.com/ellmos-ai/ellmos
  Family-level text-based operating-system documentation and infrastructure context.

## Memory pillar (.MEMORY)

ellmos memory is a family of three modules, importable as a dependency by any OS layer instead of each one rebuilding its own memory/task system:

- usmc: https://github.com/ellmos-ai/usmc
  Curated session/core memory; the facade and entry point of the memory system (push model).
- gardener: https://github.com/ellmos-ai/gardener
  Memory supplier: organic absorb/decay plus a federated cross-source FTS5 index via observe() that cites back to source (pull model). Two equally current roles: a memory module when composed into the .MEMORY pillar, and a minimal, searchable LLM-native OS core with a single SQLite table when run standalone.
- task-master: https://github.com/ellmos-ai/task-master
  Standalone SQLite task module, kept separate from knowledge memory. Zero dependencies.

## MCP servers

- CodeCommander: https://github.com/ellmos-ai/ellmos-codecommander-mcp
  Code analysis, refactoring, JSON repair, encoding repair, format conversion and diff tooling.
- FileCommander: https://github.com/ellmos-ai/ellmos-filecommander-mcp
  Local file management, process sessions, async search, OCR, archives and safe delete tooling.
- Clatcher: https://github.com/ellmos-ai/ellmos-clatcher-mcp
  Utility server for file repair, format conversion, duplicate detection and batch operations.
- n8n Manager: https://github.com/ellmos-ai/n8n-manager-mcp
  MCP access to n8n workflow automation.
- ControlCenter: https://github.com/ellmos-ai/ellmos-controlcenter-mcp
  MCP profile dashboard, capability bundles and policy audits.
- Homebase: https://github.com/ellmos-ai/ellmos-homebase-mcp
  Local LLM memory, knowledge, persistent state, routing, tests and orchestration tools.
- ServerCommander: https://github.com/ellmos-ai/ellmos-servercommander-mcp
  Server operations via MCP: HTTP health checks, access-log analysis, deployment dry-runs and mail status.
- Blender Use: https://github.com/ellmos-ai/ellmos-blender-use-mcp
  Headless Blender asset QA: background script runs and FBX reimport verification. No add-on, no TCP port.
- Open Compute: https://github.com/ellmos-ai/open-compute-mcp
  npm launcher for the open-compute MCP server: model-agnostic computer-use (screen capture, safety-gated actions, Windows UIA) over MCP.

## Reusable modules

- clutch: https://github.com/ellmos-ai/clutch
  Provider-neutral model routing with auto-learning and budget zones.
- connectors: https://github.com/ellmos-ai/connectors
  Portable messaging connectors for AI agents: Telegram, Discord, Signal, WhatsApp, Home Assistant and Webhook; framework-decoupled via SecretAdapter.
- MarbleRun: https://github.com/ellmos-ai/MarbleRun
  Autonomous multi-round agent chains for Claude Code.
- swarm-ai: https://github.com/ellmos-ai/swarm-ai
  Parallel LLM coordination patterns including stigmergy, consensus and specialist workflows.
- n8n Workflow Manager: https://github.com/ellmos-ai/n8n-workflow-manager
  Local-first GUI, REST API and CLI for n8n workflow creation, viewing and sync.
- ellmos-stack: https://github.com/ellmos-ai/ellmos-stack
  Self-hosted AI research and knowledge stack with Ollama, n8n, Rinnsal and KnowledgeDigest.
- skills: https://github.com/ellmos-ai/skills
  Reusable agent skill library for development, research, education and infrastructure.
- build-your-users-mind: https://github.com/ellmos-ai/build-your-users-mind
  Recipe for any AI agent to build a self-improving theory-of-mind model of its user from interaction logs.
- open-compute: https://github.com/ellmos-ai/open-compute
  Model-agnostic computer-use core for Claude, OpenAI CUA and mock backends with normalized coordinates, canonical actions and a central safety gate.
- project-docs-template: https://github.com/ellmos-ai/project-docs-template
  Agent-ready project documentation template with START/STATE/TODO/DONE, workflows, lightweight tooling and LLM-friendly project memory.
- anonymizer: https://github.com/ellmos-ai/anonymizer
  Local-first pseudonymization of documents: fail-closed NER, authenticated key encryption, all-or-nothing publication contract.
- report-forge: https://github.com/ellmos-ai/report-forge
  Domain-neutral core for anonymizable report pipelines: extract sources, build a schema-bound LLM prompt, fill a Word template.
- web-scraper: https://github.com/ellmos-ai/web-scraper
  Standalone web scraper extracted from BACH: get/links/forms/headers/extract/screenshot, Python stdlib core with optional extras and an SSRF guard.
- coma: https://github.com/ellmos-ai/coma
  COMAS (COMmunication for Autonomous Subagents): agent lifecycle layer (spawn, file protocol, status polling). Zero dependencies, standard library only.
- companion-for-agy: https://github.com/ellmos-ai/companion-for-agy
  PTY-based wrapper that captures agy (Gemini CLI) responses via ANSI color extraction for Claude Code, Codex, and CI pipelines.
- memoryhooker: https://github.com/ellmos-ai/memoryhooker
  Connects local memory sources to coding-agent lifecycle hooks (no network).
- roshambo: https://github.com/ellmos-ai/roshambo
  Multi-agent coordinator: serializable leases + outcome memory on CockroachDB for agents that do not know about each other.
- roshambo-starmap: https://github.com/ellmos-ai/roshambo-starmap
  Shared sky/map companion to roshambo.
- sqlite-transit-sync: https://github.com/ellmos-ai/sqlite-transit-sync
  Local-first SQLite synchronization through verified snapshots and configurable row-level merge policies. Zero dependencies, Python 3.10+.
- system-gap-master: https://github.com/ellmos-ai/system-gap-master
  Serverless sync yard for multi-machine, multi-agent AI setups (slot rule, gated daily ritual, bootstrap runbook).
- ticket-master: https://github.com/ellmos-ai/ticket-master
  Cross-platform multi-provider ticket router and triage console for CLI coding agents: files structured tickets, scores them, and routes to Claude Code, Codex and Gemini.
- lock-master: https://github.com/ellmos-ai/lock-master
  Portable zero-dependency multi-agent file-lock system: exclusive and team locks (LOCK*.txt) with scopes, expiry, stale cleanup and cloud-sync support.
- llm-note: https://github.com/doc-bricks/llm-note
  Related doc-bricks module extracted from BACH patterns: local-first notes and notebook inboxes for LLM agents with SQLite, plain-text notebooks, CLI/Python API, six locales and a standalone skill.
- knowledgedigest: https://github.com/file-bricks/knowledgedigest
  Related file-bricks module: local-first knowledge base with LLM preprocessing to ingest, structure and query documents without cloud dependencies; core module of ellmos-stack.

## Stacks

Manifest-driven compositions (schema ellmos.stack.v2) — no code copies, only declared components.

- stacks: https://github.com/ellmos-ai/stacks
  Catalog and shared manifest schema for every stack in the ellmos-ai family.
- ellmos-stack: https://github.com/ellmos-ai/ellmos-stack
  Self-hosted local-first AI research stack: Ollama, n8n, Rinnsal memory, KnowledgeDigest (file-bricks) search, and Docker Compose automation.
- agent-ops-stack: https://github.com/ellmos-ai/agent-ops-stack
  Manifest-driven composition of the local agent-ops ecosystem: ticket-master, lock-master, system-gap-master, build-your-users-mind, skills, and an MCP control plane for CLI coding agents.

## Domain tools

- law-checker: https://github.com/ellmos-ai/law-checker
  Source-grounded AI first-look legal assessments for German law (Erstorientierung, not a substitute for a lawyer): statute registry, embodiment agents, web-verified case law.
- steuer-assistent: https://github.com/ellmos-ai/steuer-assistent
  Offline-first local worksheet for German employee income-related expenses (Werbungskosten): guided collection, plausibility checks, no cloud upload.
- worksheet-generator: https://github.com/ellmos-ai/worksheet-generator
  Generates structured, ICF-aware worksheets for pedagogical and therapeutic professionals from a local LLM agent, rendered to Markdown, HTML or DOCX.

## Media and content workflows

- ai-media-editor: https://github.com/ellmos-ai/ai-media-editor
  Local AI media-editing orchestrator for video, audio and podcasts with faster-whisper or WhisperX transcription, Scribe-compatible transcript JSON, video-use cuts, Hyperframes motion graphics and agent-guided edits.
- workflowhooker: https://github.com/ellmos-ai/workflowhooker
  Git hook and webhook event trigger automation engine.

## Agents

Telephone agents built on CALL-E (https://github.com/CALLE-AI/call-e-integrations). Each takes a spoken task off a person's hands and reports what actually happened, including nonresponse.

Agent roles also ship inside modules: ticket-master (TICKET-MASTER router/triage agent), task-master (TASKSOLVER/TASKWRITER/MAINTAINER roles), build-your-users-mind (decision avatar), law-checker (statute embodiment agents plus first-look assessment agent), ai-media-editor (agent-driven creative editing). Agents shipped as skills in the skills library: research-agent (PubMed/arXiv research pipeline, from BACH's ResearchAgent) and dev-soft-agent (automated software-development pipeline), both under https://github.com/ellmos-ai/skills. Agent infrastructure (coordination, orchestration, lifecycle): swarm-ai, roshambo, MarbleRun, coma.

- hungrycall: https://github.com/ellmos-ai/hungrycall
  Sequential call cascade for food delivery, table reservations and pickup: ranks candidates against the user's boundaries, calls them one after another and exits early on the first success. Ships the generalized calling cascade pattern separately from the food use case.
- ringedingeding: https://github.com/ellmos-ai/ringedingeding
  Multi-recipient response aggregator: asks one question to several people in your own circle and merges the replies into a single result — intersected availability for a date, or the leading tendency with countervoices and reasons. Never turns dissent into a false consensus.
- researchcall: https://github.com/ellmos-ai/researchcall
  Standardized telephone survey runner with methodological honesty: eight gated stations from research question to reporting, randomized sampling and time windows, one call per person by default, and nonresponse reported without collapsing distinct outcomes. Fully local dry run by default.

## Competition entries

Projects built for public hackathons and competitions. Listed as entries; no placement claimed.

- hungrycall, ringedingeding, researchcall: CALL-E "Your Code Is Calling" (Devpost, 2026)
  Three telephone agents on the CALL-E platform; see Agents above.
- roshambo: https://github.com/ellmos-ai/roshambo — CockroachDB x AWS Hackathon (2026-07)
  Multi-agent coordinator: serializable leases and outcome memory on CockroachDB, with an MCP interface.
- roshambo-starmap: https://github.com/ellmos-ai/roshambo-starmap — CockroachDB x AWS Hackathon (2026-07)
  Evidence artefact of the accompanying multi-vendor swarm run: a replayable field record of 27 agents coordinating through roshambo across two machines.
- build-your-users-mind: https://github.com/ellmos-ai/build-your-users-mind — agent recipe entry
  Recipe for any AI agent to build a self-improving theory-of-mind model of its user from interaction logs.
- bach: https://github.com/ellmos-ai/bach — agent OS entry
  The full local-first LLM operating system: memory, handlers, skills, agents and GUI.

## Organization and maintenance

- .github: https://github.com/ellmos-ai/.github
  Organization profile, shared community health files and this llms.txt context file.
- ellmos-tests: https://github.com/ellmos-ai/ellmos-tests
  B/O/E evaluation framework for SKILL.md-based LLM operating systems and agent hubs.
- clirec: https://github.com/ellmos-ai/clirec
  Human-readable GUI demonstration recordings for CLI and agent workflows.
- recludos-legacy: https://github.com/ellmos-ai/recludos-legacy
  Archived predecessor to BACH; kept as a historical reference.

## Audience

- **LLM agent developers** building local-first pipelines who need Python/SQLite infrastructure, pluggable memory, tasks, connectors, and chain automation.
- **MCP integrators** adding file management, code analysis, n8n automation, or server ops tools to Claude Desktop or other MCP clients.
- **AI researchers** studying multi-agent coordination, swarm patterns, and consensus protocols using local tools without cloud dependencies.
- **LLM and RAG systems** that need a concise, canonical organization summary for retrieval-augmented generation or agent-context loading.

## Search Phrases

```
ellmos-ai
XLLM-OS LLM operating system
local-first AI agent infrastructure
BACH LLM-OS Python SQLite
Rinnsal lightweight LLM memory
gardener minimal LLM OS
Model Context Protocol MCP server
CodeCommander FileCommander MCP
n8n workflow automation MCP
agent memory tasks connectors
multi-agent swarm coordination Python
local-first LLM agent notes llm-note
ellmos-tests LLM OS evaluation
ai-media-editor local AI media editor
local transcription faster-whisper WhisperX video-use
Hyperframes AI video podcast editor
AI agent messaging connectors Telegram Discord WhatsApp
build-your-users-mind theory of mind user model
open-compute model agnostic computer use
Claude OpenAI CUA normalized action schema
project-docs-template agent ready project docs
START STATE TODO DONE LLM project memory
usmc gardener task-master memory pillar
curated session memory organic cross-source FTS5 index SQLite task tracking
task-master standalone SQLite task module
ellmos-stack agent-ops-stack manifest-driven stack composition
law-checker German legal AI first-look assessment
steuer-assistent German Werbungskosten offline tax worksheet
anonymizer pseudonymization fail-closed NER GDPR
report-forge anonymizable report pipeline docx schema
worksheet-generator ICF pedagogical therapeutic worksheets
web-scraper SSRF guard get links forms extract screenshot
clirec CLI agent GUI demonstration recordings
CALL-E telephone agent Your Code Is Calling
hungrycall calling cascade food delivery reservation
ringedingeding multi recipient response aggregator
researchcall standardized telephone survey nonresponse
roshambo CockroachDB serializable lease multi-agent coordinator
ticket-master lock-master agent operations tooling
```

## Last-checked: 2026-08-05